问题

I had previously installed Traefik 2 via Helm on my K8s cluster at home. 我之前在家里的K8s集群上通过Helm安装了Traefik 2。

I don't remember the specific version but I know it was some version of Traefik 2. 我不记得具体版本，但我知道它是Traefik 2的某个版本。

My cluster is actually several Raspberry Pis where I installed Kubernetes with K3s. 我的集群实际上是由几台树莓派组成的，我在上面使用K3s安装了Kubernetes。

For many months it had been working great. 多个月来，它一直运行得很好。

When I created Services with Ingress objects, Traefik would detect this and act like a load balancer, exposing the service to a specific host name. 当我创建带有Ingress对象的服务时，Traefik会检测到并像负载均衡器一样运行，将服务暴露给特定的主机名。

Today I ran a job that was apparently a little too resource-intensive because it evicted the Traefik pods for OOM errors. 今天我运行了一个工作任务，显然占用了过多资源，因为它因OOM错误而驱逐了Traefik的Pod。

Even after killing the large job, the Traefik pods didn't come back up on their own, so I uninstalled Traefik (via helm uninstall) and then used Helm to install the latest version (2.9.10). 即使杀死了这个大的任务，Traefik的Pod也没有自行恢复，所以我卸载了Traefik（通过helm uninstall），然后使用Helm安装了最新版本（2.9.10）。

In particular I used the following command to install it (where I substituted 1.2.3.4 with my real IP): 特别是我使用了以下命令来安装它（其中我用我的真实IP替代了1.2.3.4）：

helm upgrade --install traefik --set dashboard.enabled=true --set rbac.enabled=true --set="service.externalIPs={1.2.3.4}" --set="additionalArguments={--api=true,--log.level=INFO,--providers.kubernetesingress.ingressclass=traefik-internal,--serversTransport.insecureSkipVerify=true}" traefik/traefik

However, although Traefik is now running without any obvious errors, the Ingress Controller functionality doesn't seem to work. 但是，尽管Traefik现在运行正常且没有明显错误，但Ingress控制器功能似乎不起作用。

When I look at the Traefik dashboard under Services, I see api@internal, dashboard@internal, noop@internal, ping@internal, and prometheus@internal... but nothing else. 当我查看服务下的Traefik仪表板时，我只看到api@internal，dashboard@internal，noop@internal，ping@internal和prometheus@internal...但没有其他内容。

Previously I saw services listed there on the dashboard to correspond to each Ingress object I had defined. 以前，我在仪表板上看到了列出的服务，对应于我定义的每个Ingress对象。

I can post the full YAML of any of my Ingress objects if desired, but suffice it to say that all of them have the kubernetes.io/ingress.class: traefik annotation defined. 如果需要，我可以发布任何Ingress对象的完整YAML文件，但可以说所有这些对象都定义了kubernetes.io/ingress.class: traefik注释。

Has something changed with the most recent version of Traefik? 最近版本的Traefik有什么变化吗？我使用的helm upgrade --install命令与多个月前使用的相同。但正如我所说，这次Ingress控制器功能似乎不再起作用。

helm upgrade --install traefik --set dashboard.enabled=true --set rbac.enabled=true --set="service.externalIPs={1.2.3.4}" --set="additionalArguments={--api=true,--log.level=INFO,--providers.kubernetesingress.ingressclass=traefik-internal,--serversTransport.insecureSkipVerify=true}" traefik/traefik

答案1

得分: 1

因为您在helm安装命令的额外参数中提供了--providers.kubernetesingress.ingressclass=traefik-internal，所以它"不起作用"，而traefik文档中对于这个参数指出，只有具有注释kubernetes.io/ingress.class: traefik-internal的Ingress对象才会被处理。

然而，您的Ingress对象的注释是kubernetes.io/ingress.class: traefik，这就是为什么它"不再起作用"的原因。

要解决这个问题，您可以运行带有调整参数的helm升级命令：

helm upgrade --install traefik --set dashboard.enabled=true --set rbac.enabled=true --set="service.externalIPs={1.2.3.4}" --set="additionalArguments={--api=true,--log.level=INFO,--serversTransport.insecureSkipVerify=true}" traefik/traefik

这将使traefik处理所有没有注释、具有空值或值为traefik的Ingress对象。

如果您想要更加严格，那么可以使用以下命令：

helm upgrade --install traefik --set dashboard.enabled=true --set rbac.enabled=true --set="service.externalIPs={1.2.3.4}" --set="additionalArguments={--api=true,--log.level=INFO,--providers.kubernetesingress.ingressclass=traefik,--serversTransport.insecureSkipVerify=true}" traefik/traefik